This proposal seeks support for an epidemiological study of persons with mental disorders that co-occur with alcohol and/or drug abuse disorders. We would design and administer a structured clinical interview in a nationally representative sample of 12,000 persons aged 15-54. The data collection would have several unique features aimed at increasing accuracy of reporting and completeness of coverage. Interviewer observations, informant reports, a nonrespondent survey within the larger survey, and a unique matching of 1990 Census household data with data records for nonrespondents (made possible by special arrangement with the Census Bureau) would all be used in a comprehensive nonrespondent study. This set of activities would yield estimates for the total household population of the United States rather than for the 70-75% of cooperative respondents typical in general population surveys. Analysis aims would include the following: (1) We would estimate comorbidity of alcohol abuse and/or drug abuse disorders with a wide range of mental disorders; (2) We would investigate the natural history of comorbidity; (3) We would study the consequences of comorbidity for course of illness; (4) We would study risk factors for first onset and recurrence, evaluating the effects of primary disorders on secondary disorders through direct pathways and as modifiers of other risk factor influences; (5) We would study service utilization, service needs, and barriers to utilization. The survey would be carried out with the highest standards of quality control. A comprehensive set of risk factors would be assessed with the best available measures. ISR data collection methodologists would carry out special pilot studies to enhance respondent comprehension and interest. Data postprocessing would use state-of-the-art methods of multiple imputation for item nonresponse and model-based weighting for unit nonresponse. Data analyses would use modern methods of survival analysis for time-varying exogenous variables to study both risk factors for and consequences of comorbidity. Significance would be assessed with efficient complex variance estimation procedures. All major analyses would be subjected to sensitivity analyses to evaluate the effects of unit and item nonresponse.